Bug 3688 – gettext-0.11.1:uniname.c: compiler bug or impossible asm statements

• Product: Gentoo Linux
• Status: RESOLVED INVALID

[Description RoX ... 2002-06-13 08:41:58 UTC]

Hello,  
Write you for a bug (it's the place to do so)  
  
I try to make the script/bootstrap.sh  
with this make.conf :  
USE="3dnow aalib acl alsa apm arts atlas avi berkdb crypt cups dga directfb doc dvd  
encode esd evo fbcon flash gd gdbm ggi gif -gnome gphoto2 gpm gtk guile icc icc-pgo  
imap imlib innodb ipv6 java jpeg kde lcms ldap libg++ libgda libwww mbox mikmod mitshm  
mmx motif mozilla mpeg mysql nas ncurses nls nocardbus odbc oggvorbis opengl oss pam  
-pcmcia pda pdflib perl pic plotutils png -pnp postgres python qtmt quicktime readline  
-ruby samba sdl slang snmp socks5 spell sse ssl svga tcltk tcpd tetex tiff truetype X  
xface xml xml2 xmms xv zlib"  
GENTOO_MIRRORS="http://www.ibiblio.org/gentoo"  
ARCH="x86"  
CFLAGS="-march=i686 -O3 -pipe -fstrength-reduce -fthread-jumps -fcse-follow-jumps  
-fcse-skip-blocks -frerun-cse-after-loop -frerun-loop-opt -fgcse  
-fexpensive-optimizations -fschedule-insns -fschedule-insns2 -funroll-loops "  
CXXFLAGS="$CFLAGS"  
MAKEOPTS="-j3" 
  
My computer :  
2 * CPU :  
model name : AMD Athlon(tm) MP 1500+  
cpu MHz    : 1333.389  
cache size : 256 KB  
fpu        : yes  
fpu_except : yes  
flags      : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36  
mmx fxsr sse syscall mmxext 3dnowext 3dnow  
Mobo :  
 Tyan S2466 Tiger MPX  
Memory :  
 2 * 512MB  
Disk :  
 FUJITSU         MAN3184M SUN18G         1502  
 
GCC: 
Master portage # gcc -v 
Reading specs from /usr/lib/gcc-lib/i486-pc-linux-gnu/2.95.3/specs 
gcc version 2.95.3 20010315 (release) 
Master portage # gcc --version 
2.95.3 
 
  
Environnement :  
 Run in chroot from a Slack 8.1beta with kernel 2.4.19pre5-xfs-crypto-some_other_stuff 
 (kernel works fine for more than 1 month now) 
  
 
and I obtain this :  ( cf the end )  
Did someone have this type of error or do I need to report that to GCC team ?  
  
make[2]: Leaving directory `/var/tmp/portage/gettext-0.11.1/work/gettext-0.11.1/lib'  
Making all in libuniname  
make[2]: Entering directory  
`/var/tmp/portage/gettext-0.11.1/work/gettext-0.11.1/libuniname'  
gcc -DHAVE_CONFIG_H -I. -I. -I.. -I.. -I../lib -I../lib    -march=i686 -O3 -pipe  
-fstrength-reduce -fthread-jumps -fcse-follow-jumps -fcse-skip-blocks  
-frerun-cse-after-loop -frerun-loop-opt -fgcse -fexpensive-optimizations  
-fschedule-insns -fschedule-insns2 -funroll-loops  -c `test -f uniname.c || echo  
'./'`uniname.c  
gcc -DHAVE_CONFIG_H -I. -I. -I.. -I.. -I../lib -I../lib    -march=i686 -O3 -pipe  
-fstrength-reduce -fthread-jumps -fcse-follow-jumps -fcse-skip-blocks  
-frerun-cse-after-loop -frerun-loop-opt -fgcse -fexpensive-optimizations  
-fschedule-insns -fschedule-insns2 -funroll-loops  -c `test -f test-names.c || echo  
'./'`test-names.c  
test-names.c: In function `test_inverse_lookup':  
test-names.c:253: fixed or forbidden register 0 (ax) was spilled for class AREG.  
test-names.c:253: This may be due to a compiler bug or to impossible asm  
test-names.c:253: statements or clauses.  
test-names.c:253: This is the instruction:  
(insn 139 138 136 (parallel[  
            (set (reg:SI 2 %ecx)  
                (truncate:SI (lshiftrt:DI (mult:DI (sign_extend:DI (reg:SI 3 %ebx))  
                            (sign_extend:DI (reg:SI 1 %edx)))  
                        (const_int 32 [0x20]))))  
            (clobber (scratch:SI))  
        ] ) 232 {smulsi3_highpart} (insn_list/c:REG_DEP_ANTI 134 (insn_list/j/c 128  
(insn_list/j/c 138 (nil))))  
    (expr_list:REG_DEAD (reg:SI 1 %edx)  
        (expr_list:REG_UNUSED (scratch:SI)  
            (nil))))  
make[2]: *** [test-names.o] Error 1  
make[2]: *** Waiting for unfinished jobs....  
make[2]: Leaving directory  
`/var/tmp/portage/gettext-0.11.1/work/gettext-0.11.1/libuniname'  
make[1]: *** [all-recursive] Error 1

[Comment 1 Seemant Kulleen (RETIRED) 2002-06-13 18:34:34 UTC]

Bart, are those USE settings ok for bootstrapping?  Somehow I thought not..

[Comment 2 Bart Verwilst 2002-06-14 00:12:55 UTC]

Try cutting down on your CFLAGS.. I think this will solve this bug... 

[Comment 3 RoX ... 2002-06-14 03:43:15 UTC]

yeah, i know, it would pass if I cut off the CFLAGS, 
but If I choose gentoo it's in order to optimize all my progs 
(what's done with this CFLAGS). 
 
So I want to know if I could compile thsi with my CFLAGS. 
 
It's maybe a GCC bug  
( I know that GCC team don't want to put more time  
on bugs fixing on high-optimisation levels)   
but I want to know, who can fix this or if I must fix it alone. 
 
thx 

[Comment 4 Bart Verwilst 2002-06-21 11:55:40 UTC]

well, if you know what flag causes the b0rkage, i will filter it out.. i can't filter out all flags though :) 

[Comment 5 Bart Verwilst 2002-10-02 04:53:24 UTC]

Since there is no answer anymore here, i'm closing this. Lemme know what flag b0rks your ebuild, so we can filter that one out. Thanks!